Errors in the growth model used in stock assessments of Antarctic krill, specifically higher rates of growth than occur in reality, could inadvertantly lead to over-exploitation of the krill population and cause impacts on krill predators. We review the three classes of models available for modelling krill growth – the von Bertalanffy class, the moult-cycle class based on measures of the duration of the intermoult period and the instantaneous growth rate at moult, and the energetics class, based on the energetics of growth. We present a new model which is a combined energetics moult-cycle model that can flexibly take account of regional and interannual variation, including time trends, of temperature and food supply. The parameterisation for the energetic-moult-cycle model provides results consistent with the general expectations for krill growth models as well as having the flexibility to be generally applied under spatially and temporally varying conditions. In the context of assessments of precautionary yield for krill in CCAMLR, we recommend that this new model be incorporated into those assessments. The current uncertainty in many of the parameter values could be included in the Monte Carlo population projections that are used to determine yield, particularly in relation to the future trends in food supply of krill.